Yes, that definitively is one promising approach (and I hope that we would make a rough impact analysis before deciding on it and implementing it, once the structures and data are there).

I wonder if there are other approaches that are somehow more subtle. But I cannot express what I am looking for, and maybe yours is already sufficiently close to optimal.


On Fri, Sep 16, 2016 at 11:00 AM Daniel Kinzler <daniel.kinzler@wikimedia.de> wrote:
Am 16.09.2016 um 19:41 schrieb Denny Vrandečić:
> Yes, there should be some connection between items and lexemes, but I am still
> hazy about details on how exactly this should look like. If someone could
> actually make a strawman proposal, that would be great.
>
> I think the connection should live in the statement space, and not be on the
> level of labels, but that is just a hunch. I'd be happy to see proposals incoming.

My thinking is this:

On some Sense of a Lexeme, there is a Statement saying that this Sense refers to
a given concept (Item). If the property for stating this is well-known, we can
track the Sense-to-Item relationship in the database. We can then automatically
show the lexeme's lemma as a (pseudo-)alias on the Item, and perhaps also use it
(and maybe all forms of the lexeme!) for indexing the item for search.  So:

  from ( Lexeme - Sense - Statement -> Item )
  we can derive ( Item -> Lexeme - Forms )

In the beginning of Wikidata, I was very reluctant about the software knowing
about "magic" properties. Now I feel better about this, since wikidata
properties are established as a permanent vocabulary that can be used by any
software, including our own.

--
Daniel Kinzler
Senior Software Developer

Wikimedia Deutschland
Gesellschaft zur Förderung Freien Wissens e.V.

_______________________________________________
Wikidata mailing list
Wikidata@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikidata